Managing Interactions
Interactions are the foundation of Pulse. Each interaction represents a discrete, measurable user action — with a clear start event, an expected end event, and latency thresholds that define what "good" looks like.
Creating an Interaction
- Navigate to Interactions in the left sidebar
- Click Create Interaction
- Fill in the configuration:
| Field | Description |
|---|---|
| Name | A descriptive name (e.g., Checkout → Payment, Login → Home) |
| Start Event | The span or event name that begins the interaction |
| End Event | The span or event name that completes the interaction |
| Low Threshold | Latency below this is "Excellent" (e.g., 200ms) |
| High Threshold | Latency above this is "Poor" (e.g., 1500ms) |
- Click Save
Pulse immediately begins evaluating all matching sessions against these thresholds.
Interaction Detail Page
Clicking on an interaction opens its detail view with four key metrics:
Apdex Score
A 0–1 score measuring user satisfaction with the interaction's performance. Calculated by comparing each session's duration against your configured thresholds. See Core Concepts for the full formula.
Error Rate
The percentage of interaction instances where the user did not reach the end event. This captures abandoned flows, crashes mid-interaction, and navigation away from the journey.
Latency Percentiles
- P50 — The median experience. Half of users complete the interaction faster than this.
- P95 — 95% of users complete within this time. Reveals the upper end of "normal."
- P99 — Tail latency. Highlights worst-case performance affecting a small but real segment.
User Experience Breakdown
Each interaction session is classified into four categories:
| Category | Meaning |
|---|---|
| Excellent | Below low threshold, no errors |
| Good | Within threshold range, low error rate |
| Average | Mid-range performance, moderate issues |
| Poor | Above high threshold or high error rate |
Segmenting Interaction Data
The interaction detail page automatically segments results by:
- Device & OS — Which devices have the worst Apdex?
- Region & Network — Is latency spiking in a specific geography or on a specific ISP?
- App Version — Did the latest release degrade this interaction?
- Custom Attributes — Filter by any attribute your SDK sends (user tier, experiment group, etc.)
Editing and Managing Interactions
- Edit thresholds — Adjust low/high thresholds as your performance expectations evolve
- Pause/Resume — Temporarily disable tracking for an interaction without deleting it
- Delete — Remove interactions that are no longer relevant
Best Practices
- Name interactions by user intent, not technical implementation (e.g., "Add to Cart" not "POST /api/cart")
- Set thresholds based on user expectations, not server capabilities — a checkout should be faster than a data export
- Start with your revenue-critical flows — checkout, payment, onboarding, and login
- Review thresholds quarterly as your product evolves
Next Steps
- Core Concepts — Deep dive into Apdex, error rate, and UX categorization formulas
- Anomaly Detection — How Pulse flags when an interaction degrades
- Alerting — Get notified when interaction health drops